First measurement day not included when started after midnight?

[JeroenHasselaar]

For the first two participants of my study, I installed the accelerometer to start measuring at 06:00 AM on the day the participant started to wear it. When analyzing their data in GGIR, I notice that this day does not appear in the daysummary file in part 5. After noticing this, I installed the accelerometer to start measuring one day earlier (which works well). However, is there still a way to include the first day of these participants?

[tarensanders]

The most common reason for this is because your wear time criteria (includedaycrit and includedaycrit.part5) is not met on the first day. You could try reducing it and running on those participants to see if that resolves the issue. But there are lots of other parameters that can cause this too. If you post your full GGIR() call it would be easier to help.

[JeroenHasselaar]

Many thanks for your reply. Unfortunately, reducing includedaycrit and includedaycrit.part5 did not resolve the issue. Below is my full GGIR() call. Would it be possible to have a look at it and identify potential other parameters that can cause the issue? Your help is greatly appreciated.

`GGIR(datadir=RawData_folder,
outputdir = output_folder,
do.report = c(1,2,3,4,5),

 # Part 1
 #======================
 do.enmo = TRUE,
 do.neishabouricounts = FALSE, # if TRUE, calculates ActiLife counts. To use these counts, the 'acc.metric' parameter should be set to 'NeisabouriCount_vm', 'NeisabouriCount_x', 'NeisabouriCount_y' or 'NeisabouriCount_z'.
 printsummary = TRUE,
 rmc.sf = 100, # sample frequency
 windowsizes = c(1, 900, 3600),
 
 # Part 2
 #======================
 data_masking_strategy = 1, # a specific number of hours should be masked from the start and/or end of the recording, specified by 'hrs.del.start' and 'hrs.del.end'
 includedaycrit = 0, # minimum number of valid hours in calendar day
 do.imp = FALSE, # do not impute missing values because accelerometers were not worn during the night
 hrs.del.start = 0*24, 
 hrs.del.end = 0, # based on visual inspection of the plot to check data quality, after running GGIR the first time
 max_calendar_days = 12, 
 
 qlevels = (1440-MX_metrics) / 1440, # Calculates MX-metrics using the values specified earlier with the variable 'MX_metrics'
 ilevels = c(), # Would allow for calculation of narrower intensity distribution. If left empty then the intensity levels will not be extracted.
 winhr = MX_metrics, # Looks for the least and most active _consecutive_ number of X hours in a LX and MX-analysis respectively.
 iglevels = 1, # Calculates the intensity gradient. By providing a single value, bins of 25 mg are used.
 
 # Part 3 and 4
 #======================
 sensor.location = "Hip",
 acc.metric = "ENMO",
 HASPT.algo = "NotWorn", # sleep algorithm for non-wear during the night
 ignorenonwear = FALSE, # FALSE is recommended setting when HASPT.algo = "NotWorn" is used

 # Part 5
 #======================
 timewindow = "MM", # summary statistics are derived from Midnight to Midnight. This setting is necessary to be able to use 'qwindow'.
 segmentWEARcrit.part5 = 8/24, # Default. Fraction of qwindow segment expected to be valid.
 segmentDAYSPTcrit.part5 = c(0.9, 0), # Default. Also used for qwindow. Includes all segments that overlap for at least 90% with waking hours.
 includedaycrit.part5 = 0, # 8 valid hours during the waking day are required

 mvpathreshold = c(), # Blank means that MVPA is not estimated; otherwise, numbers in this vector specify over which thresholds (in g units) MVPA should be calculated
 threshold.lig = 40, # default
 threshold.mod = 69, # Hildebrand et al.(2022)
 threshold.vig = 259, # Hildebrand et al.(2022)
 boutdur.in = c(1,10,30,60), # Calculates inactive bouts lastig 30 min., 10-30 min. and 1-10 min. of duration
 boutdur.lig = c(1,10), # Calculates light physical activity bouts lastig >10 min. and 1-10 min. of duration
 boutdur.mvpa = 1, # Calculates MVPA bouts of a duration longer than 1 minute.
 frag.metrics = "TP", # Calculates Transition Probability

 # Part 6
 #======================
 part6CR = FALSE # No circadian rhythm analysis is performed. (This would calculate interdaily variability and interdaily stability)
 )

`

[tarensanders]

Does the first day show in the part2 files? What do the n_valid_hours and n_nhours columns look like on the first day (in the part 2 day summaries)?

[JeroenHasselaar]

Yes it does appear in part2 I see now! N valid hours and N hours for this day are 9.25 and 18, respectively. For the other days, there are multiple days with lower values for N valid hours than 9.25, but it is the only day with N hours equal to 18 (this value is 24 for the other days).

[tarensanders]

Yeah hopefully that is intuitive. I think with timewindow = "MM", the first day would be dropped because there is no starting midnight but I am not completely confident in this.

In any case, the solution you've found (starting the recording a day prior to wear) is our standard operating procedure.

If all you are after is PA metrics for the first day, you may be able to pull these from Part 2, FYI.

[JeroenHasselaar]

Okay that is clear. Thank you for your help!

[vincentvanhees]

Yeah hopefully that is intuitive. I think with timewindow = "MM", the first day would be dropped because there is no starting midnight but I am not completely confident in this.

If the recording starts after midnight and before 4am, GGIR estimates the wake-up time to help ensure this day can still be included in the part 5 output for the WW window even though it will be skipped for sleep results.